Composition de services basée sur les relations sociales entre objets dans l’IoT Service composition based on social relations between things in IoT

Abstract

With the rapid development of service-oriented computing applications and social Internet ofthings (SIoT), it is becoming more and more difficult for end-users to find relevant services to create value-added composite services in this big data environment. Therefore, this work proposes S-SCORE (Social Service Composition based on Recommendation), an approach for interactive web services composition in SIoT ecosystem for end-users. The main contribution of this work is providing a novel recommendation approach, which enables to discover and suggest trustworthy and personalized web services that are suitable for composition. The first proposed model of recommendation aims to face the problem of information overload, which enables to discover services and provide personalized suggestions for users without sacrificing the recommendation accuracy. To validate the performance of our approach, seven variant algorithms of different approaches (popularity-based, user-based and item-based) are compared using MovieLens 20M dataset. The experiments show that our model improves the recommendation accuracy by 12% increase with the highest score among compared methods. Additionally it outperforms the compared models in diversity over all lengths of recommendation lists. The second proposed approach is a novel recommendation mechanism for service composition, which enables to suggest trustworthy and personalized web services that are suitable for composition. The process of recommendation consists of online and offline stages. In the offline stage, two models of similarity computation are presented. Firstly, an improved users’ similarity model is provided to filter the set of advisors for an active user. Then, a new service collaboration model is proposed that based on functional and non-functional features of services, which allows providing a set of collaborators for the active service. The online phase makes rating prediction of candidate services based on a hybrid algorithm that based on collaborative filtering technique. The proposed method gives considerable improvement on the prediction accuracy. Firstly, it achieves the lowest value in MAE (Mean Absolute Error) metric and the highest coverage values than other compared traditional collaborative filtering-based prediction approaches.